Analysis of 3832 Patients over 15 Years Confirms Unusual, Genome-Wide Copy Number Variation in Patients with Developmental Disability-Autism

1. Introduction

Sequencing of the human genome,¹ followed by the ability to characterize genomic variation across individuals and populations,² has documented novel DNA changes in patients with intellectual-developmental disabilities³⁻⁸ and comorbid behaviors like autism.⁹⁻¹⁰ Chromosomal microarray analysis (CMA) enables genome-wide screening for subtle DNA excess/deficiency¹¹⁻¹² and adds copy number variants (CNVs) to the larger aneuploid segments detected by routine karyotype.¹³ These microdeletions or duplications, often created by non-allelic homologous recombination among repetitive elements, are now recognized as common individual variations that in some situations predispose to a unique type of genomic disease¹⁴ with special management implications.⁴ Microarray has improved diagnostic sensitivity from the 5 million base pair (5 Mb) resolution of high-resolution karyotyping to that of a few or several hundred nucleotides, here discussing the latter resolution obtained by microarray alone rather than the former that applies to combined microarray-exome sequencing.¹⁵ Diagnostic yield has increased from the 2-3% by karyotype to 20-25% by microarray analysis,³⁻⁹ recognition of over 50 new disorders¹³ and any chromosome imbalance endorsing microarray as a first-line test for patients with cognitive and neurobehavioral differences.¹¹⁻¹²

This report adds another survey of chromosome-microarray analysis findings to many in the literature,^{3−10, 15−19} describing 15 years of results from a single laboratory. It highlights some of the problems in transitioning from microscopic characterization of a few hundred microscopic heteromorphisms¹³ to qualifying the significance (pathogenic,¹⁸ uncertain,⁶ not or benign¹⁶)^{2, 12, 20−23} of thousands of submicroscopic CNVs^{8, 13, 24} using their prevalence in ostensibly normal versus affected people,²⁵⁻²⁶ size,¹⁷ and gene content.²⁰ Overlap of size between CNVs qualified as pathogenic versus of uncertain significance is here demonstrated as one indication of interpretative difficulties, the much higher diagnostic yield of microarray alone over karyotyping in children with disabilities and the finding of 180 novel CNVs as predispositions affirming prior endorsements of utility. Distribution of CNVs over all chromosomes with an average imbalance of almost 1% of the genome in these patients highlights their potential for early therapy-invoking screening for neurodisabilities including epilepsy or autism.

2. Materials and Methods

2.1. Sample Sources

Array-comparative genomic hybridization (chromosomal microarray analysis or CMA) was introduced at the Texas Tech Cytogenomics Laboratory in January 2009, supplementing long-standing cytogenetic testing. This study includes all non-oncology samples processed through June 2024. Whole-exome sequencing (WES) data from July 2022 to August 2025 were integrated with corresponding CMA results for comparative diagnostic analysis. Most samples were obtained from medical center and private practice physicians across the West Texas region extending from Amarillo to El Paso North-South and roughly to Abilene/Midland Odessa to the East. All testing was performed under the supervision of the authors (VJT and SC) within the Department of Pediatrics at Texas Tech University Health Science Center, Lubbock. All analyses adhered to CAP and CLIA guidelines, with compliance verified through biennial inspections.

2.2. Sample Analyses

Prior to 2009, the laboratory operated as a standard cytogenetic laboratory, performing chromosome analyses using conventional G-banding methods as previously described.²⁷ Microarray analysis was introduced in 2009 as an added test, with karyotyping remaining the preferred initial analysis. Beginning in 2016, automated karyotyping of heparinized peripheral blood or fibroblast cultures was followed by DNA extraction and microarray analysis using validated protocols.²⁸ Microarray platforms evolved as follows: from 2009 to 2015, the EmArray Cyto 6000 44k chip (Agilent Technologies, Santa Clara CA) with BlueGenome software (Illumina/Novagene, Sacramento CA) from February 2016 to June 2024, the Cytosure Constitutional 60k CGH array (Oxford Gene Technologies, Begbroke UK) with Cytosure® Interpret software; and from June 2024 onward, the Agilent GenetiSure Cyto CGH 60k Array with CytoGenomics software (Agilent Technologies, Santa Clara CA). Arrays were processed according to manufacturer specifications, including hybridization, washing, and scanning on Agilent SureScan systems. CNV detection and interpretation followed ACMG consensus guidelines,^{11, 20} incorporating probe-level quality metrics, log2 ratio thresholds, assessment for copy-neutral loss of heterozygosity, and correlation with ClinVar,²³ DECIPHER,²⁹ and the Database of Genomic Variants.³⁰ Twenty-four CNVs classified as likely pathogenic in the 2016-24 results were grouped with pathogenic findings for reporting consistency.

3. Databases

All laboratory samples and reports were entered into a deidentified database under IRB approval (Texas Tech IRB-FY2025-128, exempt status granted March 12, 2025). The database was constructed from pre-existing laboratory systems used for patient-physician reports, billing, and accounting. Data were stored on password-protected computers with access restricted to IRB-authorized laboratory personal and approved student/faculty researchers, in compliance with CAP, CLIA, and HIPPA standards. Identifiable information was limited to laboratory accession number, sex, and age (grouped into 1- year intervals for children under five and five-year intervals for older patients). Karyotypes were entered in ISCN (2020,2022, 2024 format as applicable), and all CNVs were documented by chromosomal band location and base-pair coordinates, along with classifications as benign, VUS, or pathogenic according to ACMG guidelines.^{11−12, 21−23}

4. Database Analysis and Statistics

Data tallies were generated using Microsoft Excel, employing search, find, and sort functions to organize CNV counts, classifications (pathogenic, VUS, benign). Descriptive statistics, including means, standard deviations, and frequency distributions, were calculated using Excel’s standard formulas. Statistical significance for group comparisons was assessed using two-tailed t-tests for continuous variables (e.g., mean CNV size per patient) and N-1 chi-squared tests for categorical variables (e.g., diagnostic yield differences between cohorts).³¹ A p-value <0.05 level was considered statistically significant.

5. Results

Table 1. summarizes 15 years of microarray analysis, results by patient in the left columns and by numbers of copy number variants to the right. The 16,138 total CNVs in 3832 total patients (data row 3) documented from 2009 to 2024 yielding an overall average of 4.2 variants per patient that will be further detailed in Fig. 1. There were 1524 patients receiving microarray results from 2009 to 2015 (data row 1, left), 264 having at least one CNV qualified as pathogenic, 58 or 1202 having only CNVs qualified as variants of uncertain significance (VUS) or benign (left columns, Table 1). Thus 264 or 17% of patients were judged to have a diagnostic result by microarray, the corresponding numbers 485 or 21% for the 2016-24 patients (row 2) and 749 or 20% for the combined groups (row 3). The number of CNVs in these patients is shown in the right columns of Table 1, those qualified as pathogenic comprising a respective 8.6, 4.2 and 5.2% of the 2009-15, 2016-24, and combined patient groups with benign CNVs being the vast majority (94% for the combined group, data rows 1-3). The higher percentage (8.6%) of pathogenic variants in the 2009-15 group, unaccompanied by more pathologic diagnoses (17%), suggests some selective entry of those variants or fewer defined syndromes when the 2009-15 database was compiled.

2009-15, 2016-24, and combined (darkly shaded) results of microarray analysis are shown in the top 3 rows, total patients (pts, 1^st data column) and CNVs (6^th column) converted to 100% in the lightly shaded 2^nd and 7^th columns so proportions of diagnoses or variant (var) qualifications could be indicated in columns 3-5 and 8-10.¹Only 3408 patients and their associated 15257 variants were designated by sex, 424 patients in the 2009-2015 database entered as sex unknown; 2many patients had testing indications of developmental (DD) or intellectual (ID) disability and birth defects combined; “Other” includes non-disability diagnoses like Ehlers-Danlos syndrome, male infertility due to azoospermia, recurrent pregnancy loss. CNV or var, copy number variant; Dx, diagnostic; path, pathogenic; testind, test indication; VUS, variant of insignificance;*significantly larger (p <0.05) fithan percentage for years 2009-15 or 2016-24; **significantly larger than percentage for males; #significantly less than percentage for patients with DD-ID-birth defects testing indication; ^&significantly larger than percentage for patients with DD-ID-birth defects testing indication.

The far-right column gives total patients, those to the left 2015-24 cases by 2-year intervals, years 2015 and 2024 having data for the last and first six months respectively; 494 patients had microarray results only (138 with culture failure) and 1976 had chromosome results (data row 4); the second column assigns letters to total cases (a), those with a pathogenic copy number variant (CNV-c), those with an abnormal karyotype (e) and so forth, the third column explaining the percentages; VUS, variant of uncertain significance. *significantly higher or # significantly lower (p <0.05) biannual case percentage than that of all cases in the far-right column

Diagnoses by combined karyotype and microarray analysis as compared to those by either alone are shown in Table 2, these results taken from years 2015-24 when chromosome results were listed in the laboratory database. Total patients are in the far-right column, the 2470 patients with microarray or chromosome results equaling the 2308 for years 2016-24 in Table 1 plus another 162 patients from 2015 (including 119 with chromosome results and 3 with culture failure). Biannual patient numbers are listed in the 5 columns to the left of total patients, numbers and percentages of these total patients (top row) for various categories listed in successive rows (right columns, Table 2). Thus the patients given a diagnosis by karyotype or microarray (aCGH) are in the second row, their number (b) equaling the sum of patients with a pathogenic CNV (c) plus those with abnormal karyotypes (e) after subtracting patients where the pathogenic CNV confirmed (f--i. e., large 21 duplication paralleling karyotype of trisomy 21), clarified (g—i. e. microdeletion at breakpoints of an apparently balanced translocation), or added an unrelated CNV to the chromosomal diagnosis (h—i. e. a pathogenic 17q12 microdeletion in a patient with trisomy 21 karyotype). Patients diagnosed by karyotype and microarray analysis combined totaled 520 or 21% of total patients overall, with similar percentages for the two-year intervals from 2015-16 to 2023-24 (Table 2, second row).

Note that chromosome analysis alone gave diagnoses in 267 or 14% of patients overall (far right column, fifth data row) microarray adding diagnoses in 7% of patients for the 21% overall diagnosis rate. Microarray did add diagnostic information to cases with abnormal karyotypes in a higher 12% of patients, providing identities of marker chromosomes (too small for accurate band identification) or demonstrating that microduplications/deletions do or do not accompany translocations that by karyotype appear balanced. The ability of karyotyping to show polyploidy, translocations. and levels of mosaicism not demonstratable by microarray¹³ suggests consideration of a tiered approach to chromosome-microarray testing (see Discussion). Note also the increased yield that might come from qualifying more VUS-qualified CNVs (3.1-4.3% of totals in Table 2, last row) as pathogenic using point scores.²⁰

Looking at the data for 2-year intervals shows some progression in diagnostic sensitivity as one goes from the 2015-16 (19%) to 2023-24 (25%) patients (second row, Table 2). Increased percentages of diagnosis are also shown for pathogenic microarray (19 to 25%) and abnormal karyotype results (12 to 19%) as one proceeds from those biannual cases (third and fifth data rows, Table 2), The additional pathogenic microarray findings unrelated to the abnormal karyotype vary from 0 to 6.5% over these years (8^th data row), the percentage having a benign or VUS CNV accompanying the abnormal karyotype showing a 2.1 to 8.7% increase (9^th data row). Similar percentages for CNVs adding diagnostic information and for VUS microarray results over the years (bottom 2 rows of Table 2) shows that resolving VUS qualifications as benign versus pathogenic has not improved with time.

Figure 1. Number per patient and average sizes of CNVs found in 3832 patients from 2009 to 2016. A, number of variants per patient for total microarray results 2009 to 2015 (9-15), 2016 to 2024 (16-24), and 2009 to 2024 (9-24), the combined 9-24 results then tallied by benign, pathogenic, or VUS qualification. B, distribution of CNV average sizes, smaller ones (left) in kilobases (Kb), larger ones in megabases (Mb) on right. Microdeletions and duplications are grouped together in the indicated CNV size intervals.

Figure 1. Number per patient and average sizes of CNVs found in 3832 patients from 2009 to 2016. A, number of variants per patient for total microarray results 2009 to 2015 (9-15), 2016 to 2024 (16-24), and 2009 to 2024 (9-24), the combined 9-24 results then tallied by benign, pathogenic, or VUS qualification. B, distribution of CNV average sizes, smaller ones (left) in kilobases (Kb), larger ones in megabases (Mb) on right. Microdeletions and duplications are grouped together in the indicated CNV size intervals.

In Fig. 1A, the number of variants per patient is shown for the 2009 to 2015 microarray analyses (most having 2 CNVs per patient), those from 2016 to 2024 (most having 4 CNVs per patient), and the combined group (most having 2 CNVs per patient). The greater number of CNVs per patient for the 2016 to 2024 group reflects the change from the older EMArray 44,000 DNA probe chip to the newer Cytosure Constitutional 60,000 DNA probe chip (see Methods). Separate plotting of variants by qualification from the combined group shows the expected maximum of 2 per patient for benign variants (94% of all CNVs from Table 1), numbers of pathogenic or VUS variants declining sharply from a majority of 1 CNV per patient (Fig. 1A, left).

Fig. 1B shows the average size of CNVs, microdeletions and duplications grouped together, ranging from ~100 base pair microduplications or deletions to the ~5 megabase DNA segment size that would be detected by routine karyotype to the 30 megabase+ aneuploid segments that would be found in patients with traditional trisomy/monosomy syndromes. Note again the majority of benign variants (numbers divided by 60) and the substantial numbers of pathogenic variants (numbers divided by 2), the former peaking at sizes of 11-99 Kb in contrast to the latter at 2001-4984 Kb or about 2-5 Mb. A challenge for interpreting microarray results is shown by the overlap of CNV sizes for those qualified as VUS (~20 in the 2001-4984 Kb range, ~15 in the 10-19.9 Mb range) and those qualified as pathogenic (right side of Fig. 1B). Even a few CNVs qualified as benign had sizes in the 1001-4984 Kb range, reaffirming that size alone cannot be used to qualify CNVs as disease-contributing.^11,20

Diagnoses made by routine karyotype, sometimes accompanied by fluorescent in situ hybridization (FISH), and/or by chromosome microarray analysis are shown in Table 3. Large CNVs that confirmed aberrations visualized by karyotype (full or mosaic trisomies, translocations producing large aneuploid segments) are not listed in the right columns, smaller ones that confirmed, clarified, or added novel microarray to karyotype findings as enumerated in Table 2 are listed. There were 61 chromosome disorders involving specific chromosomes or bands diagnosed by karyotyping, another 18 like triploidy not listed (see Table 3 legend). All of these were confirmed when microarray analysis was performed, 55 pathogenic CNVs (33 microdeletions, 22 microduplications) found in 3 or more patients (31 of them in 5 or more) shown in Table 3. Among these were 26 microdeletions and 17 microduplications recognized as known disorders.

Also correlating with the pathogenic CNVs listed in Table 3 are the lists of specific pathogenic or benign CNVs in S1 Table of the Supplemental Information, benign variants selected for display by 1) being in the same chromosome region as a pathogenic variant, and 2) recurring in more than 15 patients as a benign variant. S1 Table thus gives an overview of disease-associated human copy number variation, exact band loci and average CNV sizes provided in contrast to Table 3 where overlapping pathogenic variants are grouped (e. g., the one 4q32.1/.2 and three 4q32.2q35.2 microduplications in the lower left of S1 Table are grouped as four 4q32.1q35.2 microduplications in the right column of Table 3). S1 Table lists 62 CNVs (37 microdeletions and 25 microduplications) that met strong association criterion (in three or more patients) and 36 (24 microdeletions) occurring in five or more patients. Another 32 (14 microdeletions) almost meet criteria by occurring in two patients and 86 (39 microdeletions) in one, foreshadowing the large number of disorders with subtle chromosome imbalance that will be delineated by microarray analysis.

• ^11.Not listed are patients with balanced translocations (11), marker chromosomes (4), pericentric inversions (2), or triploidy (10). Only the 54 microdeletions/duplications found in 3 or more patients are listed, recognized microdeletion/duplication syndromes in larger type and bolded; aCGH, array-comparative genomic hybridization or microarray analysis; CduC, cri-du-chat; DG, DiGeorge; Dx, diagnosis; PW/A, Prader-Willi/Angelman; SM, Smith-Magenis; Sx or S, syndrome; Transloc., translocation; W, Williams; WH, Wolf-Hirschhorn.

There are equal numbers (90 each) of loci with pathogenic microdeletions or microduplications in S1 Table, the former occurring in 362 patients and the latter in 222. Average sizes (including 1 CNV for each locus rather than multiple ones) are 6439 Kb for microdeletion CNVs and 6895 Kb for microduplication CNVs. Loci with 40 or more benign CNVs (shaded in S1 Table) included 23 with microdeletions and 26 with microduplications, their average lengths being a respective 263 and 277 Kb. Again the overlap of sizes is shown with CNVs qualified as pathogenic including a deletion of 329 Kb (12q23.1) and duplications of 267 Kb (2p22.1), 338 (4q12), and 387 (4q32.1), larger CNVs qualified as benign including a deletion at 5q13.2 (1400 Kb) and duplications at 5q13.2 (1100 Kb), 10q11.21 (1300 Kb), 18q21.1 (1100 Kb), and Xp22.2p21.3 (1300 Kb).

All of the pathogenic CNVs have loci in common with benign CNVs except for the 3 at 11q24.1q25 (middle right of S1 Table). A slight majority 35 of the 62 pathogenic CNVs occur in regions where there are 20 or more benign microdeletions and duplications, less than half (29 CNVs) sharing loci where there were 100 or more benign variants of either type. Of the loci with frequent benign CNVs, the 23 with multiple benign microdeletions accompanied by 15 with pathogenic microdeletions, the 26 with multiple benign microduplications accompanied by 13 with pathogenic microduplications. Locations of larger pathogenic CNVs did not significantly associate (p <0.05) with those having frequent smaller benign CNVs, indicating genesis of larger and thus likely pathogenic CNVs can be a primary event rather than further amplifications of benign CNVs.

Adding to the overview of CNVs detectable in the human genome by medical laboratory analysis is the data in S2 Table of the Supplementary Materials, first looking at the distribution by chromosome of benign microdeletions and duplications (columns A-0). CNVs qualified as VUS or pathogenic (216 or 836 variants in Table 1) are not included in these columns because they are a small proportion of the 15,083 benign variants; they are included in columns P-Q where the non-overlapping CNV nucleotide lengths and their fraction of chromosome length are listed. Although CNVs occur on all 24 chromosomes, we see from the conditional formatting in columns C and H that chromosomes 6, 8, 17, and 22 have more microdeletion CNVs (present in a respective 1224, 1016, 601, and 704 of the 3832 patients analyzed) and that chromosomes 8, 14, 17, 22, and X have more microduplication CNVs (present in a respective 1618, 947, 910, 836, and 1346 of patients analyzed).

Multiplying these numbers of CNVs (columns C and H) by their average length (columns D, I) gives the number of variant nucleotides in kilobases, which can then be divided by the 3832 patients to give the amount of nucleotide variation per patient (E, J). Dividing nucleotide variation by chromosome length (repeated in columns A, R for reference) gives the nucleotide variation per patient as a proportion of each chromosomal DNA strand (columns F and K). Once average CNV sizes are incorporated, there are more deleted nucleotides on chromosomes 1, 5, 8, and 14 (respective amounts of 33, 60, 21, and 14 Kb) and more duplicated nucleotides on chromosomes 8, 14, 17-15, 22, and X (respective amounts of 94, 98, 40-44, 39, and 50 Kb). Dividing these numbers of variant nucleotides by chromosome length then gives larger percentages of deleted nucleotides for chromosomes 1, 5, 8, 16, and Y (respective percentages of 0.013, 0.033, 0.014, 0.016, and 0.012% in column F) and of duplicated nucleotides for chromosomes 8, 14, 17-15, 22, and Y (respective percentages of 0.064, 0.092, 0.051-0.044, 0.078, 0.032,and 0.042). The bottom row totals of S2 Table indicate that the average patient in this study had 256,000 nucleotides of deleted DNA amounting to 0.19% of their genome (columns E, F), 542,000 nucleotides of duplicated DNA amounting to 0.56% of their genome (columns J, K).

Overall amounts of benign-qualified copy number variation are shown to the right of S2 Table, 798,000 nucleotides of copy number change (deletion plus duplication) occupying greater percentages of chromosomes 5, 8, 14-17, 21, X, and Y and constituting 0.72% of the average patient genome (column O). The excess of pathogenic microdeletions (362 versus 222 microduplications) shown in S1 Table is reversed for benign CNVs that total 5984 microdeletions and 9099 microduplications (bottom row of S2 Table).

Column N of S2 Table shows that the ratio of total benign microdeletions to microduplications varies widely by chromosome, numbers 5 and 6 having substantially more microdeletions (darker red conditional formatting), numbers 1, 2, 10, and 13 having slightly more or nearly as many, the other eighteen chromosomes having more duplications with numbers 9, 11, 12, and 21 having substantially more (dark blue formatting). When ratios of CNV nucleotides per patient per chromosome length are tallied, chromosomes 1-3, 5, 9, 11, 13, 18, and 20 have more deletions per strand of DNA with chromosomes 8, 14, 17, 22, and X having an excess of duplications. The contrasting excess of pathogenic microdeletions compared to that of benign duplications fits with the greater impact of deleted DNA known from chromosome aberrations and autosomal recessive conditions.

A different perspective on CNV genotype/clinical phenotype association is suggested by the data in Table 4, showing that the presence of a (typically larger) pathogenic CNV increases the average number of CNVs in the average patient. Recall from Fig. 1B that the different chips used for microarray analysis in the 2016-24 population versus the 2009-15 patients gave more average variants, numbers of CNVs in the presence or absence of a pathogenic variant given for each patient group and all patients. The 3.0, 6.4, and 5.3 mean variants per patient when a pathogenic CNV is present are significantly more than the 2.3, 5.1, and 4.0 when absent at the p <.0.001 level. Also increasing average numbers of CNVs per patient is the presence of a large chromosome aberration found by routine karyotype, raising the oft-discussed possibility that DNA imbalance decreases regulatory constraints on genes within and outside of aneuploid intervals (see Discussion).

CNV, 0001. level.